Let $A=\begin{pmatrix} 1 & 1 & 1 \\ 1 & 1 & 1 \\ 1 & 1 & 1 \end{pmatrix}$
Its eigenvalues are $\lambda=0,0,3$.
Find two pairs of orthonormal eigenvectors for $\lambda=0$ and
show that $P=x_1 x_1^{T}+x_2 x_2^{T}$ is the same for both pairs.
(These eigenvectors make orthogonal matrix $Q$, and $A=Q\lambda Q^{-1}$)

I found first pair $x_1=\frac{1}{\sqrt{2}}(-1,1,0), x_2=\frac{1}{\sqrt{2}}(-1,0,1)$
and second pair $x_1=\frac{1}{\sqrt{6}}(2,-1,-1), x_2=\frac{1}{\sqrt{2}}(0,1,-1)$
but then $P$ of these two pairs are not same.
Where am I wrong?


I solve it. Put $x_1=\frac{1}{\sqrt{2}}(-1,1,0), x_2=\frac{1}{\sqrt{6}}(1,1,-2)$ in the first pair.
I'm not sure there is any particular formula for this question,
but I just put all possible eigenvectors here.
Thanks for your answers and comments.

  • $\begingroup$ Once again: the new pair of vectors $\,x_1,x_2\,$ you found are not orthogonal. $\endgroup$ – DonAntonio Dec 11 '12 at 3:15
  • $\begingroup$ Actually, I'm not sure what "orthogonal" means. Before normalizing these two eigenvetors, they are orthogonal ($x_1=(-1,1,0), x_2=(1,1,-2)$ and then after normalizing them, they are not orthogonal? ($x_1=\frac{1}{\sqrt{2}}(-1,1,0), x_2=\frac{1}{\sqrt{6}}(1,1,-2)$) $\endgroup$ – niagara Dec 11 '12 at 3:41
  • $\begingroup$ I calculated it and they are orthogonal. Their inner product produces zero. $\endgroup$ – niagara Dec 11 '12 at 3:44
  • $\begingroup$ Now they are... $\endgroup$ – DonAntonio Dec 11 '12 at 10:18

The first pair is not orthonormal. The inner product is $1/2$.

Since any nonzero constant multiple of an eigenvector is also an eigenvector, we take $x_1=(-1,1,0)$ and $x_2=(-1,0,1)$. Then $\langle x_1,x_2 \rangle = -1\cdot -1 + 1 \cdot 0 + 0 \cdot 1 = 1 \neq 0$.

  • $\begingroup$ You're right! But then how can I find the first pair? I have to try all possible eigenvectors? $\endgroup$ – niagara Dec 11 '12 at 3:00
  • $\begingroup$ I don't think this answer is very helpful. It should be a comment rather than an answer, but @niagara you should orthogonalize the given matrix which then will give you the desired vectors. $\endgroup$ – diimension Dec 11 '12 at 6:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.